European Central Bank Gauges Lenders’ Risks From AI

The ECB is asking lenders for further details on their loans to areas such as data centers, Bloomberg News reported Monday (Feb. 23), citing sources familiar with the matter.

At the same time, the ECB is holding workshops to see how banks are employing artificial intelligence, the report added. Bloomberg notes that these efforts highlight the way regulators are recognizing AI’s potential to shake up the banking industry.

Banks and private credit companies around the world have invested trillions into building out AI, whether that means development companies, data centers or energy supply, the report added.

The central bank declined to comment when reached by PYMNTS.

The ECB workshops center on things like business models, governance, and risk management. A source told Bloomberg that at least one bank understood the ECB’s wish to signal the need for caution when it comes to lending to sectors like data centers.

These efforts are happening at a time when AI agents are becoming more deeply ingrained into compliance queues, cash management dashboards and payment routing engines.

It’s a shift that “marks the first real test of whether financial institutions trust AI with operational authority,” PYMNTS wrote earlier this month.

“Unlike earlier generative AI tools that responded to prompts, agentic systems can plan, reason and execute multistep workflows across systems with limited human intervention,” the report said. “Financial institutions are embedding these systems into compliance, treasury, risk and payments infrastructure, signaling a shift from automation pilots to production-grade deployment.”

Agentic AI workflows are starting to reconfigure how regulated financial work gets done, especially in anti-money-laundering (AML) and know your customer (KYC) investigations, according to a recent Thomson Reuters analysis.

Rather than analysts manually collecting data across sanctions lists, corporate registries and adverse media databases, AI agents can autonomously compile, reconcile and document findings while maintaining an audit trail fit for regulators.

“It’s a move from task automation to workflow orchestration, where AI coordinates across systems rather than assisting in isolated steps,” according to the report.

Last year, PYMNTS examined the rise of cognitive banking, which refers to the combination of AI-driven inferencing and pattern recognition and permissioned data (transactions, financial behaviors, linked accounts), letting banks shift from reactive servicing to proactive guidance.

PYMNTS Intelligence research found that nearly three-quarters of surveyed bank customers wished for greater personalization, and that embedded conversational artificial intelligence could win back 72% of bank customers by offering that tailored experience.